nyukat/breast_cancer_classifier

Implements dual-pathway classification combining image-only and image-heatmap models in PyTorch, processing standardized 16-bit mammograms across four standard views (CC and MLO) with per-breast benign/malignant probability outputs. Includes a patch-classifier-based heatmap generator that produces pixel-level attention maps (0-1 range) corresponding to benign and malignant regions, which are concatenated with raw images for the enhanced model pathway. Provides both exam-level (four-view) and single-image inference modes, with TensorFlow alternative and pretrained weights available.

886 stars. No commits in the last 6 months.